Scores by kind of claim
Each kind of claim is scored on its own, out of 100: the share that held up, with the range it likely falls in. Fewer resolved claims give a wider range.
-
Factual claims
Statements about the present or the past, checked against the record: did they hold up?
80 74 86 -
Predictions
Statements about events outside the person's control, checked once they fell due: did they come true?
Insufficient record: 4 predictions resolved, 5 needed for a score.
-
Promises
Commitments to the person's or organisation's own future action, with something to deliver: was it delivered?
85 74 94 -
Guidance
A company's own figures for its results over a period, checked against what it then reported.
Insufficient record: 0 guidance claims resolved, 5 needed for a score.
All types combined: 81 (likely 76 to 86, high confidence), from 86 claims: 70 factual claims, 16 promises. It mixes kinds of claim in different proportions for different people, so compare people kind by kind.
Mark Zuckerberg's claims
151 claims · page 2 of 2
“We've basically gotten to like 18 or 19 categories of harmful things that people do”
Mark Zuckerberg, Meta CEO, interviewee 18 April 2024 Mark Zuckerberg - Llama 3, Open Sourcing $10b Models, & Caesar Augustus (Dwarkesh Podcast)
Meta has identified about 18 or 19 categories of harmful activity on social media and built AI systems to detect them.
Unresolvable The public record could not settle it.
The second judge would not resolve this claim; the verdict shown is the first judge's.
The evidence is the person's or their organisation's own records.
Said for an organisation they lead or work for; it counts on both records.
What would count as true or false
Judge as of 2024-04-18. True: Meta's own records (Community Standards, transparency/Community Standards Enforcement Reports, or official newsroom posts) or credible reporting from around that date show that Meta had defined roughly 18–19 categories of harmful content or behaviour (17–20 is close enough, given the claim's 'like 18 or 19') and used AI/automated systems to detect content in those categories. Mostly true: the categories and AI detection are confirmed but the count is somewhat off (about 15–16 or 21–23), or AI detection covers most but not all of the categories. Mixed: one part holds (the category count or the AI detection) but the other cannot be confirmed or clearly fails. Mostly false: the count is far off (for example under 12 or over 30), or the categories exist but Meta did not use AI detection for them as a set, so the claim overstates a real practice. False: the evidence shows Meta had no such categorised framework, or did not use AI to detect harmful content.
Evidence
- Integrity Reports, First Quarter 2024 | Transparency Center Meta (Transparency Center) · 4 December 2024 Own record
“An increase in actioned content for Violence and Incitement on Instagram due to updates to our detection technology that improved identification of hostile speech.”
- Integrity Reports, First Quarter 2024 | Transparency Center Meta (Transparency Center) · 4 December 2024 Own record
“An increase in actioned content for Suicide and Self Injury on Instagram due to accuracy improvements in our detection technology.”
- Integrity Reports, First Quarter 2024 | Transparency Center Meta (Transparency Center) · 4 December 2024 Own record
“Prevalence of violating content remained relatively consistent across a wide range of violation areas.”
- Integrity and Transparency Reports, First Quarter 2023 Meta (about.fb.com newsroom) · 17 May 2023 Own record
“For example, we’ve added more metrics like data for appeals and restores as well as data for more policy areas over time.”
Reasoning
The evidence shows Meta updated its "detection technology" in two harm areas on Instagram [E1][E2]. It does not say this technology is AI or automated, or that it covers the categories as a set. None of the evidence gives a number of harm categories [E1-E4], so neither part of the claim is clearly confirmed.
Cite this claim
Mark Zuckerberg, “Meta has identified about 18 or 19 categories of harmful activity on social media and built AI systems to detect them.” (said 18 April 2024). Verdict: unresolvable. Reliability API, claim c_f3d7997d10d7, method 0.1.0+3f84d14d. https://epistacklabs.com/claims/c_f3d7997d10d7
Claims without a verdict
- Unresolvable 144
- Not yet due 17
- Own records only 41
- Not researched 223
- Not checkable 1,060
Is a verdict here wrong, or do you speak for Mark Zuckerberg? Reply on the claim's own page, with the form at its foot (how replies work). Verdicts are made automatically; how verdicts and scores are made.